Toward an automatic preoperative pipeline for image-guided temporal bone surgery.

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Minimally invasive surgery often relies on time-intensive preoperative segmentation and trajectory planning.
• The temporal bone presents unique anatomical challenges for surgical planning.

Purpose of the Study:

• To develop a fully automated pipeline for segmentation and trajectory planning in temporal bone surgery.
• To improve the efficiency and reproducibility of preoperative planning for minimally invasive interventions.

Main Methods:

• A deep learning approach with shape regularization for automatic segmentation of temporal bone structures.
• Trajectory planning for linear and nonlinear interventions based on segmented risk structures.
• Evaluation on 24 patient CT datasets for planning cochlear and internal auditory canal access.

Main Results:

• The automated pipeline achieves comparable segmentation accuracy (Dice score) to semi-automatic methods.
• The approach provides more accurate organ shapes essential for precise trajectory planning.
• Publicly available source code facilitates broader adoption and further research.

Conclusions:

• Fully automated segmentation and trajectory planning for temporal bone interventions are achievable.
• The proposed pipeline streamlines preoperative planning, leading to efficient and unbiased workflows.
• This automation has the potential to enhance clinical outcomes and patient safety.